AI News for March 7, 2026 — 7min.ai

[NEWS] Amodei apologizes for leaked 'gullible' memo, argues Pentagon supply chain label is narrower than claimed: Anthropic CEO Dario Amodei confirmed the company received the Pentagon's supply chain risk designation and announced it will challenge the ruling in court.
[NEWS] Anthropic launches Claude Marketplace, letting enterprises buy partner tools through existing AI spend: Anthropic unveiled Claude Marketplace, a platform where enterprise customers can apply part of their existing Anthropic spend commitment toward third-party tools powered by Claude.
[TOOLS] OpenAI launches Codex Security, an AI agent that finds and patches code vulnerabilities: OpenAI released Codex Security (formerly "Aardvark"), an AI agent that scans code repositories, builds project-specific threat models, and validates vulnerabilities in isolated test environments.
[RESEARCH] Anthropic's Claude finds 22 vulnerabilities in Firefox, including 14 high-severity bugs: Anthropic's Frontier Red Team used Claude Opus 4.6 over two weeks to discover 22 security vulnerabilities in Firefox, 14 classified as high-severity.
[TOOLS] OpenAI launches ChatGPT for Excel powered by GPT-5.4, optimized for financial modeling: OpenAI launched ChatGPT for Excel, a beta add-in that lets users create, edit, and analyze spreadsheets through natural language.
[NEWS] Oracle and OpenAI scrap Texas data center expansion, Meta moves to fill the gap: Oracle and OpenAI abandoned plans to expand a flagship AI data center in Abilene, Texas, after negotiations stalled over financing and OpenAI's evolving needs.
[NEWS] Pentagon R&D chief details 'holy cow' moments that triggered Anthropic rift: Undersecretary of Defense Emil Michael described on the "All-In Podcast" two pivotal moments that led to the Pentagon blacklisting Anthropic.
[RESEARCH] MIT researchers compress KV cache 50x with new Attention Matching technique: Researchers at MIT developed Attention Matching, a fast compression method for the key-value (KV) cache that reduces LLM working memory by up to 50x with minimal quality loss.
[NEWS] February hit $189B in global VC funding, the largest single month ever, with 83% going to three AI companies: Global venture funding reached $189B in February, a single-month record driven almost entirely by three AI mega-rounds: OpenAI ($110B), Anthropic ($30B), and Waymo ($16B).
[NEWS] Grammarly's AI 'expert review' uses real people's identities without permission: Grammarly's "expert review" feature generates AI writing advice "inspired by" real people, including The Verge's editor-in-chief Nilay Patel, deceased professors, and dozens of tech journalists, none of whom gave Grammarly permission.

About 7min.ai · Subscribe to newsletter